Y-scores, Y-loadings, 101 y-scores – BUCHI NIRCal User Manual

Page 208: 102 y-loadings

Advertising
background image

NIRCal 5.5 Software Manual

208

NIRCal 5.5 Manual, Version A

Interpretation: difference between pretreated spectrum and reconstructed spectrum will be summed up
and squared. It is with the:

1. PC: X-PRESS 1= [pretreated spectrum-(mean spectrum + score 1 x PC 1)]2
2. PC: X-PRESS 2= [pretreated spectrum-(m.s. + score 1 x PC 1+ score 2 x PC 2)]2
3. PC: X-PRESS 3= [pretreated spectrum-(m.s. + score 1 x PC 1+ score 2 x PC 2 + score 3

x PC 3)]2 etc.


The primary principal components should be set at the point where no significant changes in the X-
PRESS value can be determined. The PC, which has no effect on the X-PRESS, normally has very
small or only noisy information (see:

Loadings

).


Here: 1-3 PCs are important.

3.18.101

Y-Scores

Description

Each spectra is placed in the n-dimensional y-score space. The position of
the spectra is given by the n-dimensional coordinate of the y-scores.

Use

For PLS2 (multi property calibration).

Method

PLS

Matrices ID

111

Tip

Compare with the Scores.

Details

Calculated for the C-Set (V-Set Y-Scores are 0).

Related Topic

Scores

,

Y-Loadings


Matrices of the final PLS calculation:

Spectra are decomposed to Scores and Loadings.

Properties are decomposed to Y-Scores and Y-Loadings.

Y-Scores are predicted from the Scores and Properties are predicted from the Y-Scores.

3.18.102

y-loadings

Description

Y-Loadings build up the base for reconstructing the property together with
the Y-Scores.

Use

For PLS2 (multi property calibration).

Method

PLS

Matrices ID

112

Tip

Details

For one C-Set property (PLS1) the Y-Loadings are constant 1.

Advertising